Jump to comment:

• Overestimate of cost of smoking to the NHS
  Christine Callum
  Published on: 13 May 2016

• Published on: 13 May 2016

  Overestimate of cost of smoking to the NHS

  • Christine Callum, Independent Consultant
  • Other Contributors:
    • Sean Boyle

  This paper addresses a number of important issues around the costs of smoking to society, and in particular to the UK National Health Service (NHS). However there are methodological issues which result in the paper overestimating the costs of smoking to the NHS. While smoking does represent a significant cost to the NHS, the estimates provided in this paper, based as they are on a mixture of very old data and parameters...

  Show More

  This paper addresses a number of important issues around the costs of smoking to society, and in particular to the UK National Health Service (NHS). However there are methodological issues which result in the paper overestimating the costs of smoking to the NHS. While smoking does represent a significant cost to the NHS, the estimates provided in this paper, based as they are on a mixture of very old data and parameters derived indirectly for a broad geographic region of which the UK is just one part and at that a rather atypical part, are of questionable validity.

  The only recent piece of UK information used by the authors refers to 2005/06 total NHS cost. The breakdown of total cost by disease and the proportion of disease-based costs attributable to smoking are derived respectively from extremely old NHS sources and from five-year old figures for a WHO region. In addition the authors assume that a method appropriate for estimating the burden of disease by age to society is also applicable to estimating health service use.

  There are 2 key elements in this study:

  identifying diseases where smoking is a factor and providing population-attributable fractions (PAFs) for these diseases; and,

  the apportionment of NHS costs to these diseases.

  We believe that the paper uses questionable data in both these areas. We use the example of the English NHS to demonstrate this. The paper suggests that the cost to the NHS in England of smoking was £4.4 billion at 2005/06 prices. We have recently calculated a figure between £2.7 billion and £3.1 billion at 2006/07 prices using a method which calculates the appropriate PAFs for England, and using more recent NHS service use and cost data (ASH 2008).

  COSTS

  Taking costs first, the paper recognises the limitation of using NHS cost data by disease category based on the proportion of costs associated with particular diseases in 1992/93. The authors even go as far as comparing their figures with the more recent NHS programme budget data, showing very large differences. But they choose to ignore this. This may be because they have misunderstood what the programme budget data cover as they wrongly state that they ‘provide(s) detailed expenditure information for primary care service…’ In fact although there are some limitations to the quality of the data, all elements of NHS expenditure are covered though till now primary care consultation costs are not yet apportioned by disease.

  It is likely that the disease-based pattern of actual NHS costs in England in 2005/06 was different from that in 1992/93 (the actual data on which the authors base their study) as there have been changes in patterns of disease, clinical practice and technology, NHS structures, and overall spend levels. This is demonstrated by the programme budget data. For example, in 1992/93 cardiovascular diseases accounted for an estimated 12.1% of costs. Using the Department of Health’s programme budget estimates of costs by disease – which apportion all hospital costs and primary prescription costs but not so far primary care consultation costs – ‘circulation problems’ accounted for 7.9% of 2005/06 NHS costs (Department of Health, 2006). If we apportion pro rata the separate category which includes all primary care consultation costs (9% of costs altogether) across disease-based categories this provides a more directly comparable figure of 8.9%. This alone reduces the estimated cardiovascular smoking cost to £1.6 billion rather than the authors’ figure of £2.1 billion. Applying the same approach to cancer and the ‘other medical’ category comprised of respiratory disease and peptic ulcer indicates an increase of £0.3 billion in cancer costs and a reduction of £0.6 billion in ‘other medical’ costs. Altogether this amounts to a £0.9 billion reduction and this alone would bring the cost down to £3.5 billion. The actual impact is likely to be considerably higher as programme budget categories include a broader set of ICD codes than in the conventional disease groups; for example, the program budget category includes all cancers and tumours. If ICD codes were matched more precisely then the implied overstatement would be considerably greater.

  POPULATION ATTRIBUTABLE FRACTIONS (PAFs)

  We also have concerns about the PAFs used by the authors in calculating costs for the UK/England. Our concern is less with the method used to estimate the PAFs than with the actual regional PAFs used; they misrepresent the UK sex and age pattern of exposure to smoking and they apply to the year 2000 rather than 2005. In estimating deaths due to smoking for example, rather than deriving UK-specific figures for 2005 to reflect the sex and age distribution of deaths by disease, the authors chose to use WHO regional PAFs. Peto et al. using the same method to estimate PAFs but based on UK data estimated that in the UK in 2000, 21.9% of male deaths and 16.1% of female deaths were attributable to smoking, whilst their overall ‘developed’ countries estimate of 22.0% male and 8.1% female deaths signified much lower female relative to male exposure. (Peto et al. 2006). More recently an estimate for England in 2005 using an alternative method indicates 22% of all male and 13% of all female deaths attributable to smoking (Information Centre 2007). The authors’ UK estimate of 27.2% male and 10.5% female deaths in 2005 based on the male- female relationship embodied in the regional PAFs appears therefore to misrepresent the UK.

  For costs furthermore the authors indicate that the PAFs for DALYs used were WHO EUR-A region (very low child and adult mortality) figures for 2002. However the separate all-age male and female PAFs used in the paper were in fact for 2000 and based upon the much broader, more varied and higher mortality ‘Developed’ region. We have attempted to correct for this by using male and female EUR-A 2000 instead of ‘Developed’ 2000 PAFs for DALYs, combining them as the authors did using the 2002 DALYs distribution for EUR-A. This has a significant impact on costs. For example, for cardiovascular disease, the paper calculates PAFs (using ‘Developed’ 2000) as 32% males, 10% females, 22% combined, whereas based on EUR-A 2000, the PAFS are 27% male, 8% female, 19% combined. The effect of this is to reduce the attributable cost by £0.3 billion (using the paper’s outdated cost proportions), or £0.2 billion using 05/06 disease- based costs. Assuming 05/06 disease costs and adjusting only the PAF for cardiovascular disease, the effect is to reduce the overall cost from £4.4 billion to £3.2 billion. However even these figures are based on PAF figures for year 2000 smoking levels; more recent figures would result in an even lower estimate of costs.

  References

  ASH (2008). Beyond Smoking Kills. London: ASH.

  Department of Health (2006). Resource accounts 2005-06. London: The Stationery Office .

  Peto R, Lopez AD, Boreham J, Thun M (2006). Mortality from smoking in developed countries 1950-2000 (2nd edition: updated June 2006).

  Information Centre (2007). Statistics on Smoking, England 2007.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top